Restoration Method and Application for Improvement of Recognition Rate of Augmented Reality Model Registration Marker in Ship Maintenance

Abstract The increased interest in smart ships also increases interest in ship maintenance, movement, and steering. Recently, a significant shift toward augmented reality (AR)-based remote maintenance methods has been witnessed. This approach allows for the visualization of the data necessary for maintenance directly in the user's field of vision, as opposed to relying on traditional voice and video maintenance methods. To utilize AR in a complex ship environment with numerous pipes, valves, and supports, it is more appropriate to use markers rather than markerless methods, which often fail to detect stable features due to visual clutter and occlusion. However, when a marker is attached to a ship, its visibility may be reduced owing to shadows caused by the equipment, corrosion, or damage, making its application difficult. Therefore, to solve this problem, we compared the GAN algorithms used for image restoration for damaged markers and selected the most efficient algorithm. Additionally, tests were conducted assuming a damaged case, and the results were evaluated.